Unanswered questions

Connecting the official Front MCP (https://mcp.frontapp.com/mcp) from Cursor as a custom remote server (confidential OAuth client, no DCR). Redirect in use: http://localhost:8787/callbackFeature access: MCP onlyResource permissions: Read + Write + Send enabled on the app as a ceiling. Client requests scope=read+write only (no send on the token). Authorize returns invalid_scope: The requested scope is not supported: read, write — immediately after the company-domain screen. The same OAuth state is reused across retries even after deleting and recreating the MCP server in Cursor. app.frontapp.com/.well-known/oauth-authorization-server lists scopes_supported: ["feature:mcp"] only. Docs say MCP scopes are read, write, send. Has anyone hit this in the beta? Happy to share more details privately with Front if you need them for a per-app fix.

We have a BYO agent that can create drafts based on @-mentions in threads. Everything was working as expected for a long time, but suddenly our backend has stopped receiving webhooks since 9 am ET this morning. We validated this across multiple conversations, also ones where we knew it was working before. 

Anyone know where I can find a Front Consultant?  Front was already in place when I started here about two years ago and has been problematic since day one.  Our Front setup is a hot mess and generates way too many helpdesk tickets for all kinds of different reasons.  I really need someone that can look at our setup, analyze it, suggest and make changes.  I’m an IT Manager but I don’t have the bandwidth to revamp our whole Front setup.  A Google search turned up nothing and Front doesn’t seem to offer such a service that I can tell.  Any insight would be greatly appreciated. Thanks

Hey everyone 👋Today we support one scorecard per workspace, and we’re working on changing that - so that you can have the flexibility of different scorecards to apply for different participants in the same conversations, with a shared criteria library to select from across your company. We’re currently exploring how to define the applicability of a scorecard, and would love your input. If you had multiple scorecards, what would be your use case / purpose for the different scorecards? How would they be ideally setup ? A scorecard per teammate group, inbox etc? 

The Front MCP to Town (town.com) connection keeps dropping, can help to troubleshoot this please? It is a critical integration for us. We need to manually reconnect it each time it drops. I’ve attached a few screenshots for reference. Thanks!  

You can now see whether a conversation is snoozed for you through the MCP server, and can snooze conversations as well. Please let us know if there are any issues using this feature!

Hi everyone!We recently expanded Copilot to a few more of our teams at their request. They saw it being used by other teams and were interested in the conversation summaries and suggested replies.As we roll this out further, I would love to hear from those who have used these features for a while: The "Honeymoon" Phase: How has the sentiment evolved? Does the team still find the summaries valuable daily, or has it become "background noise"? Accuracy: How reliable are the conversation summaries? Do your teams find them accurate for long threads, or do they still feel the need to read the entire history anyway? Suggested Replies vs. Manual Edits: Do your agents find the suggested replies "ready to send," or are they mostly using them as a skeleton and spending significant time editing them? Knowledge Base Integration: For those seeing high-quality suggested replies, do you have a Knowledge Base linked or synced? I am interested to know if the model performs well enough pulling from

Checking emails in the dark from the mobile app...so painful in light mode. We need dark mode on mobile please!

Hello,For one of our mutual customer we are seeing that conversation events are missing from /events api call.While individual event exist when queried via /conversation/[id]/events endpoints.I have also checked it’s not a private resource. Also checked the api token access it’s all correct.I can share more details if we can start a thread around api call and conversation which is missincg, since it’s one of our mutual customers data I am not posting here our public forum.Let me know how to debug this.I have already created a support ticket.

Hi all, Not sure if this is the correct location for this. I run a Technical Support team and we have inbound via email and WhatsApp as well as Zoho CRM tasks.  Zoho Desk didn’t work out for us as it would not work with custom modules of CRM. We don’t a yet have a developer.  Front looks like it may suit us but we use personal email, shared team inboxes and we also use WhatsApp groups that involve the team but also customers in those groups. Can Front work with WhatsApp groups (as well as 1 to 1 WhatsApps) and Zoho tasks?

My team and I aren’t able to send emails. We can receive them, but we get an “unable to send” error. Is this happening to anyone else? Hoping to get the resolved as soon as possible. Thank you!

It would be really helpful if emails sent automatically through Rules could have the same “Seen” tracking as manually sent emails.Our team uses open tracking to understand whether a member has viewed an outbound email, including when it was seen. This is especially useful when deciding whether to follow up and for keeping an accurate record of customer communications.At the moment, this works well for manually sent messages, but not for automated template messages or auto-replies triggered by Rules.Ideally, Rule-sent emails would show the same Seen status and timestamp in the conversation timeline as manually sent emails.

We're soon rolling out an upgraded model for Smart CSAT, and we'd like you to see it first.Smart CSAT already gives you satisfaction signal on every conversation, not just the ones customers reply to. The new model sharpens that signal considerably — it's better at reading nuance, and more disciplined about what it scores in the first place.What's new:• Knows when not to score. Internal threads, notifications, and other non-customer conversations are now recognised and left out, so your scores reflect real interactions only.• Handles open-ended conversations fairly. When a customer doesn't reply, that's no longer read as dissatisfaction.• Identifies the right customer. In conversations with several participants including other teammates who are not on front, the model better identifies external participants to measure.• A more discerning scale. Ratings are better distributed across the range, so a 5 means more than it used to.We're opening this to a small group of customers ahead of ge

Hey! I was just wondering if you could tell me more about how does the MCP work in terms of GDPR? Is our customer’s data safe/anonymised in the process? 

We’ve been listening to your feedback on being able to take action on Smart CSAT and Manual CSAT Scores. They’re now available as dynamic variables as conditions in rules to trigger actions such as tagging "low-CSAT", routing or assigning for review or triggering a customer reviews flows or follow up notes on high scoring Manual CSATs!💡 Share how you might use this in your process workflows ? 

Hey! We’re building out functionality to score every AI interaction, diagnose what went wrong, and validate changes before go-live — for your AI agents.As part of this, Smart QA & Smart CSAT help you measure quality at scale across every conversation — human and AI. Smart QA scores AI Agent performance against your quality rubric, while Smart CSAT infers customer satisfaction on AI Agent conversations.Comment on how you are measuring AI Agent quality today, if you are using any tooling or built your own, and if you’d like access to Smart CSAT + Smart QA -- for design partners you will get an extended 60 day trial (normally 14 days for Smart CSAT, 30 days for Smart QA)!Book in time with me to run through these features + preview some designs on the upcoming functionality for AI Agent quality measurement and performance tuning. ThanksHelena

Hi,We're building a BYO AI agent against the agents beta + MCP server. Draft creation works, but only once we pass an explicit channelId, and we'd like to understand the intended model before we build around our workaround.Same conversation, no existing drafts, three calls back to back:create_draft(conversationId) -> -32000: Failed to create draft replycreate_draft(conversationId, channelId=cha_AAAAA) -> -32001: You do not have access to channel cha_AAAAAcreate_draft(conversationId, channelId=cha_BBBBB) -> {"draftId": "msg_...", "channelType": "gmail"}Our reading: cha_BBBBB is one of the 13 channels list_channels returns for this agent; cha_AAAAA isn't. With no channelId, Front falls back to the conversation's own channel — also not one of the 13 — and that failure comes back as a contentless -32000 rather than the clear -32001 we get when we name the channel ourselves.What we want is simple: the agent drafts a reply on the conversation's own channel, so the draft is addressed

We are looking for an individual to work with an ongoing basis to better customize our Front setup alongside our teams to get the full use out of the platform. 

Hi Front team,Sharing a feature request for the MCP server. First, thanks for the open beta and for gathering feedback here directly.The MCP server currently has no Knowledge Base tools. I'd love to see the KB surfaced through MCP.We run a public help center on Front's KB and I'm building an AI agent to keep it healthy — specifically a periodic link audit. For each article, the agent needs to look at the outbound links to other KB articles and flag ones that are broken.Concretely, the tools I'd hope to see- read_knowledge_base_article — returns body with links as [label](url) (or a structured list of links with hrefs), plus last-modified date- list_knowledge_base_articles / list_knowledge_base_categories — for enumerationLater we also like to create drafted articles in some other workflow.So, would love to know if KB support is on the roadmap, and happy to be an early tester if useful. Thanks!

We're a sales team of just three people at a small, local growth company, and we're using a shared inbox on the front end. We're struggling to decide how many routing rules are worth creating when we're so small: should we route by customer type, by pipeline stage, or just let anyone take the conversation as it comes in? If you have a small sales team, what rule/routing configuration has worked best for you without becoming an administrative burden?

Hi,I am following your steps for connecting our Ai assistant (ChatGPT), but the Registration URL does not auto-populate. What is the Registration URL for Front’s MCP?A walk-through video showing how to connect Ai assistants would be extremely helpful (Section 2 in https://dev.frontapp.com/docs/mcp-server) 

I run an AI agent on claude that sweeps my Front every morning and surfaces what actually needs my attention, especially activity on our biggest deals. Front is the source of truth for that, so the agent reads conversations through the API (and I looked at the new MCP server too).The core problem: there's no way to programmatically get my "Open" view, the thing I look at in the UI every day. My Open is the union of my personal inbox + conversations assigned to me + conversations I'm subscribed/following. The first two are reachable. The subscribed bucket is not.Specifically:follower: search is unreliable. Searching follower:<me> returns only a handful of conversations (4 in my case) when my actual Subscribed bucket in the UI is 23. The conversation object itself correctly lists me under /conversations/{id}/followers, but I can't search or list by it. So there's no way to enumerate "everything I follow." No "Open view" union endpoint. Your own MCP docs confirm this: "There's also